Commentary
Best Practices for Video Across Platforms
- by Daisy Whitney , Op-Ed Contributor, September 11, 2015
Let’s say you’re a brand. We don’t have to convince you that you need a video strategy, right?
You have one -- surely. And since you do, let’s talk about some best practices across the leading video platforms in support of branded video content.
For starters, bear in mind some key differences among platforms. The average video length on YouTube is 12 minutes, compared to 1:28 seconds on Facebook, six seconds on Vine and about 15 seconds on Instagram, according to a white paper on multiplatform video strategy from Social@Ogilvy and video software platform Tubular Labs.
That means Facebook is best at delivering the initial burst of views and engagement, while YouTube can drive more views and engagement over a longer period of time. On Facebook, a view counts after three seconds, and many videos are played without the sound. Also, video success on Facebook is often dependent on its appearance in the news feed, while YouTube leverages its search engine
Given the realities of the platforms, Tubular and Social@Ogilvy urge creators to make the first frame of a Facebook video “quite catchy.” Brands might also want to use a title card so that users can understand the content of the video with the sound off. Since Facebook is good for quick hits in the newsfeed, it can be helpful to use videos there to jumpstart a wider campaign that reaches across other platforms.
Over on YouTube, the key is to make the first few seconds engaging so that viewers stay tuned for the entire video. Lean on YouTube influencers whenever possible, and also look for opportunities to create evergreen content that can be found through YouTube’s search.
A grabby opening is also critical on Instagram, and brands need to use hashtags effectively on that medium. On Vine, a successful strategy is often one that pairs a brand with a popular creator. Vine videos should be planned in advance so that the six-second limitation can be best utilized to tell a story, the paper says.
No differentiation here between pre-roll ads and searched content? I'd expect vastly different results. Also useful to know the differences between video that is app delivered rather than browser delivered.
User persistence numbers would give this discussion some beef. For example, for those YouTube videos that are counted "viewed", what percentage persist for 15 seconds? For 30 seconds? Etc.
The data is easy to capture when YouTube video is embedded.
Similar questions for other video distributors.
Thanks for this.
Recently a client asked me to write copy for a video and the producer insited this video (for presenting commerical real estate) be no longer than 1:30. I couldn't understand why the producer was so adamant. They must have been unaware of these differences.
Your suggestion to make the "first frame of a Facebook video 'quite catchy'" is pretty evident - it's your video's "headline."
The way we watch video is evolving pretty fast so I thank you for these "Best Practices." I'll look, consider, incorporate, and, most of all, remember, the new wave is still pretty frothy at the front....
This is not a particularly new audience measurement problem.
Simply focus on 'average minute audience' - the way TV does. With TV if someone watches one minute of a 30-minute programme they are counted as 'reached' (i.e. saw any of the programme), but they are counted only as one-thirtieth of a viewer.
So if we draw a parallel with on-line ads with (say) an average 30-second duration, they would count as a 'full viewer' if they watched all of the 30-seconds.
Of course the issue with TV is that it is a programme rating that is reported and the assumption is that ad-breaks 'rate' the same as the programme. Clearly this is rarely the case with an Australian study (unfortunately a decade old now), showing around a 5% average decline during the ad-break.
But there is an online issue as well. The 'viewing duration' is being calculated based on the server-duration. There is no knowledge of whether the ad is being seen, buffered, and whether the player is in an active browser window and tab (or a foreground app). There is also no knowledge of whether a person is there viewing the ad (though for short-form it is a pretty fair assumption they are there but they may have taken the browser out of focus) - or in fact there may be more than one person viewing.
John, the current national reporting standard for Tv in the U.S. is something akin to a "total audience" per commercial minute rating, and does not include---usually----any program content. The problem arises when you have long and short ads---mainly "15s" and "30s". Naturally there is more avoidance for the shorter ones, but, as far as I can tell, this is not accounted for.
Bearing in mind that we don't know if anybody is actually watching any of the commercials in a break, the ideal solution would be to define a threshold time frame for each commercial length---say 10+ seconds for a 15-second ad, 20+ seconds for a 30-second ad, etc. and then tally each ad separately as "viewed" or not "viewed", based on this definition. This could apply to TV as well as video. Being fair, however, this would be a real challenge for Nielsen to correlate all of the required information---especially for digital----and develop a program for churning out the numbers.
My apologies Ed and other readers.
I should have prefaced my comments with "Here in Australia...". Many other countries, (New Zealand, most Asian countries, the UK etc), report at a programme level - hence my explanation.
The origin of TV ratings was to provide data in order to assist the broadcaster to sell ads in their programmes (and accordingly they are primarily funded by broadcasters). This is why a panel-based approach is a cost-effective and efficient way to provide acceptable estimates. In essence the reporting unit is an accumulated 30 minute period (or longer).
With the commercial pressure to report on individaul ads the system does not have the granularity required to report at the 15-second or 30-second level (and was not designed - or funded - to do so). Indeed many meters only poll every few seconds, and while audio-matching is continuous, both systems tend to use some form of 'rule of dominance' to assign each minute to a single station which then 'owns' that entire minute. While that is a pretty fair assumption during a programme, it falters during ad breaks.
Having said that using audio-encoding (audio-injection rather that audio-matching) would mean that ad-content could be accurately identified. This however comes at a cost to the agency and advertiser, as well as significant changes within the ratings ecosystem and analysis software. But it can be done. [Compared to identifying ad-content on-line this would be a stroll in the park]. This issue is - who pays?
I recall hearing of a great rebuttal from a broadcaster to an agency/advertiser group meeting. The client and agency were looking at the minute-by-minute data during the ad-break and the gross audience was something like 10% less than the programme minutes. They wanted compensation either as make-goods or ratecare reductions. The broadcaster sales directors line of reasoning went something along the following lines ... so we spend tens of millions of dollars making a programme which can deliver you ten million people on a plate, you put your crappy ad in my programme and within 30 seconds you can lose one million of my audience, and you want a discount for that, why don't you make better ads.